The chainflail prototype was a built from a pre-existing root cleaner, consisting of a box-like structure supporting two rotating drums. The drums were mounted 1 m apart and were powered by two variable displacement hydraulic motors that would turn at 800–1000 rpm, depending on the rotational regime of the endothermic engine that fed them. Each drum carried 20 flails, consisting of 6 hardened chain links each. Normally, the device would be fed vertically from the top, so that the short rootstocks to be cleaned would dangle between the two drums and would be flailed until all the dirt was removed. Therefore, the first step in prototype development consisted in turning the device by 90° to enable horizontal feeding. Then, an infeed table was added, for supporting incoming tree bunches. At the other end of the flail, a metal chute was installed for holding that stem portion that had passed through the flail (Fig. 1). Two bump plates were added: one in front of the infeed table and the other at the end of the chute, for indexing tree butts and assuring accurate crosscutting of the whole bunch. Since the target log length was 4 m, the second bump plate – that at the end of the chute – was placed at 4 m from the centerline of the flail drums, so that the delimbed stem portion would extend to exactly 4 m and would be clearly visible at crosscutting. While the eventual commercial product would be fitted with its own hydraulic pump and power pack, this first prototype was designed for connecting to the hydraulic system of its transport, due to budget restrictions. All was mounted on a roll-on roll-off flat deck skip for easy transportation between sites: the total weight of the chainflail device was 3 t, including the skip that weighed 1 t itself. The whole operation was contained in a 6-axle truck-and-trailer rig, whereby the CFDD skip was loaded on the three-axle truck and the excavator tasked to feed it sat on the three-axle trailer. The excavator was a tracked 13-t model, fitted with a grapple saw for feeding the flail with whole-tree bunches, pulling out the delimbed bunches, crosscutting them at a length of 4 m and separately stack logs and tops (Fig. 2). One operator was enough to relocate and operate the whole system.
After a brief test run near the workshop in Italy, the machine was moved to Western Slovakia and tested on one of the short-rotation poplar plantations managed by IKEA Industry near Malacky, in close proximity of a major particle board factory tasked with producing a highly innovative poplar based lightweight panel. In particular, the test plantation was located near Gajary (48° 29’ 10.87” N; 16° 55’ 25.52” in WGS84), in the Morava river floodplain. Local climate was described as “warm temperate, fully humid, with hot summer climate” (Cfb) according to the Köppen-Geiger classification (Rubel et al 2017). The mean annual temperature was 11°C in the 2014–2020 interval and the average annual precipitation was 742 mm. Soil was a Mollic Gleysol, with sandy texture and groundwater levels between 1.5 and 2.0 m from the surface. The test was conducted in early March 2022. Weather during the test was consistently warm and dry, with occasional light precipitation. Air temperature varied between − 2 and + 14 C°. The plantation was a 6-year-old poplar stand established at a square spacing of 3.0 m x 2.0 m with hybrid poplar (Populus x euramericana Dode (Guinier)), clone ‘AF18‘(Heilig et al. 2021, Landgraf et al. 2020, Meyer et al. 2021).
The test machine was operated by the owner of Biomass Work Ltd., who was a qualified forestry professional with many years of experience in poplar harvesting work. He had also operated the chainflail for many years, although only in the rootstock cleaning configuration, given the absolute novelty of the new machine derived from it (Fig. 3). Nevertheless, he was quite familiar with the working principle, the expected results and the hazards of chainflail operation. Before starting the study proper, the operator worked half a day on an unmarked stack in order to perfect his routine and iron out possible difficulties. After that, the experiment proper commenced, which consisted of a time and motion study conducted over two different feedstock types: standard trees and underdeveloped trees - respectively the “strong” and the “weak” tree treatments. The former would normally yield at least one 4-m log - more often two; the latter would only yield one 4-m log, if any at all.
The experimental design was a factorial scheme where each treatment was repeated 8 times. Each repetition consisted of one pile of approximately 130 trees, in order to reflect the same batch size adopted in other similar studies conducted under the same research programme – thus achieving comparability, in case of further use of the same datasets. The chainflail would process the piles in a random order, to neutralize any potential background noise derived from machine wear or operator fatigue. To minimize the latter effect, at the end of each pile the study was halted to allow for the operator to rest, while the support team cleaned and inspected the machine for any signs of malfunction (e.g. leaks, accelerated wear etc.). Taking a brief rest pause every hour of work is a recommended practice in commercial operations, too.
The circumference at breast height of all trees in all piles was measured manually with a measuring tape and then converted into diameter at breast height (DBH), over bark. Furthermore, 6 trees covering the whole DBH distribution were destructively sampled in order to determine their total height and weight, separately for the theoretical log and chip portions (Krejza et al. 2017; Urban et al. 2015). Destructive sampling allowed estimating the relationship between DBH, total height and mass, which was used to predict the mass packed into each individual pile (Headlee and Zalesny 2019). Previous studies have shown that it is possible to build reliable allometric functions with such a small sample, when tree variability is as small as found in even-aged clonal poplar (Hartmann 2010; Hjelm 2015; Verlinden et al. 2013). Initial mass estimates were later adjusted using ad-hoc correction factors obtained by matching the estimated log and biomass yields with the actual amounts taken to the factory weighbridge. That was done separately for the log and for the chip portion obtained from each of the two treatments, in order to account for variations in log recovery that might be associated with the treatments (i.e. 4 correction factors). Moisture content was determined both at the time of destructive sampling and at the time of delivery to the factory, so as to match dry mass estimates with dry mass weighbridge data. In both cases, moisture content was determined with the gravimetric method, according to EN ISO 18134-2:2015. Mean moisture content at delivery was 55% (standard deviation = 2.7%). Depending on treatment, the ratio between factory dry mass and inventory dry mass varied from 0.75 to 1.12 with an overall average at 0.85 - meaning that the field inventory overestimated actual harvest by about 15%.
Delimbing quality was visually assessed by the factory production managers who attended the trials. Log length was regularly checked with a tape measure all along the duration of the trials. The machine was set for delimbing, not debarking.
During the test, researchers determined the time taken by the CFDD to process each individual pile, using a stopwatch accurate to the second. Both productive time and delay time were recorded (Bjorheden et al. 1995), but the latter was excluded from the study, where it was replaced by a 20% delay factor. That was done because the time spent on each pile was too short (about 1 hour) to accurately estimate delay time. The 20% increase applied to the data was consistent with the findings of previous published studies, with special reference to the harvesting of plantation forestry (Spinelli and Visser 2008). That figure was also quite close to the sum of all delays recorded during the complete study, as conducted on the 16 piles.
The pile-level time study was accompanied by a parallel cycle-level elemental time study (Magagnotti et al. 2011). That would cover more than half of chainflail cycles on each pile, where cycles were identified as the time to complete the processing of a tree bunch broken off the pile and fed to the chainflail. The total cycle then included all tasks required for turning a group of trees from the test pile into logs and biomass stacked onto their respective piles. The goal of this study component was to determine if treatment would specifically impact one or more work steps within the complete flailing task. Furthermore, the elemental time study would indicate which ones are the most time-consuming work steps and address future improvements of the prototype. This study split the complete cycle into the following work tasks (elements):
Grab = Time spent grabbing a tree bunch and indexing the trees against the bump plate. It ends when the bunch is inserted between the rotating flails (easily identified through the flail-on-wood noise). The record includes a count of trees in the bunch;
Process = Time spent delimbing the bunch and crosscutting it. It ends when the last log obtained from the bunch is crosscut. The record includes a count of the logs produced from the original bunch;
Stack logs = Time spent moving the crosscut logs onto the log stack;
Pile residues = Time spent moving the residues (tops and branches) onto the biomass pile;
Other work time = any other work time – typically clearing debris from under the infeed opening and chute etc.
The pile-level study data was used to quantify operation productivity and log yield (dependent variables) as average values, and the differences between alternative treatments (independent variables) was checked using non-parametric statistics as a safeguard against possible violations of the parametric assumptions. Given the only two treatments were being compared (“weak” vs. “strong”), a non-parametric test would not be much less informative than a standard parametric test, while being more robust – hence more reliable. In particular, the Mann-Whitney unpaired comparison test was used for this study. Since we renounced the normality assumption, centrality was represented through Medians – not Means. For all analyses, the significance level was set at α < 0.05. The analyses were implemented with the software Minitab 17, one of the most popular statistical software in the field of engineering (Okagbue et al. 2021)